Gerrymandering code – for the do-it-yourselfers
For the true hardcore…A version of the MATLAB code I used is now available. I used gerry.m for Part 1, and delegations.m for Part 2. Voting dat...
Gerrymandering notes: how many votes to take House, 1992-2012
Dear PEC Readers, we’re going casual tonight. This is for frequent visitors and those of you on the RSS feed. I’ll explain…
I’m preparing a long-form piece (for elsewhere) on the topic of partisan House gerrymandering. We’re cooking up some graphs to drive home some basic points. Your immediate reactions and critical questions will be welcome.
This graph shows what fraction of the two-party vote would have been needed for Democrats to control the House of Representatives.
The procedure was:
- Calculate the % two-party vote for all 435 districts.
- Calculate the shift in vote needed to make an outcome of exactly 218 Democratic seats.
- Add this shift to the national % Democratic vote.
The colored horizontal line segments indicate which party was in control. Generally, the out-party needs a bit more than 50% of the two-party vote to gain control. This extra barrier is an advantage for the incumbent party.
Note 1: Dealing with uncontested races is a challenge. For instance, the 2006 data point is distorted by the fact that there were 47 uncontested races won by Democrats (versus only 10 won by Republicans). Forty-seven is an unusually high number. With other definitions, this data point is more comparable to 1996-2004.
Note 2: I came into this analysis expecting the 2012 value to be unusually high because of partisan gerrymandering. It is indeed high – but it is only on a par with 2004. I am pondering if there is a problem I am missing.
This post will self-destruct in 12 hours.
Topics:
I find the control of congress color bars to be confusing when compared to the % vote line. I’d suggest the color correspond to the party that won that election (ie change color @ 2010, 2006, 1994 rather than nearly at the start of the next election). The interesting point is that the change of control has occurred when it is relatively even (except 1994), but then the gerrymander swings towards the party in control in subsequent years.
In relation to the 2012 value, it might be interesting to look at it in terms of delta from the previous election rather than absolute value, that might be a better indication of how big a shift has occurred.
To clarify, I am not looking for guidance on graphic design. There is a very good professional who will do something interesting with this. I am more looking for reactions to the substance.
I think the relatively even aspect in 1994 and 2006 might only be visible in retrospect, i.e. not predictable. It is driven a bit by the problem I mentioned re 2006. It requires more analysis.
What does the chart of number of uncontested races (D – R) look like?
It strikes me that in 2004 a Republican president won by 2 points and yet in 2012 a Democratic president won by 4 points. But they have the same voting margins for House control. Is there some coattails effect that is missing?
Adam, I am under the impression from political science literature that Presidents have downticket coattails in both directions.
Pigeon, the number of uncontested races fluctuates considerably. It’s hard to control for in a way that is easily explained. From 103rd to 113rd Congress:
R: 14, 36, 13, 57, 33, 45, 40, 10, 3, 6, 25
D: 21, 18, 10, 40, 34, 40, 31, 47, 13, 1, 21
difference: -7, 18, 3, 17, -1, 5, 9, -37, -10, 5, -4
That big dip from 9 to -37 occurs at the same time as the 2006 anomaly. Something’s in common there, not sure what since it’s quantitatively not enough by itself.
KT, Interesting. When time permits (not today, probably) I can drill into that.
I was just reminded that in Texas, redistricting is limited by the Voting Rights Act. Something like 8-10 districts are untouchable. This could explain the funny anomaly I got before that Texas redistricting doesn’t really favor Republicans in the way one might expect.
Some Body, the redistricting issue is interesting but it requires more data. For the moment, I have two cycles available to me at the moment.
re note 2: perhaps you are not ‘missing something’ in 2012, but rather have discovered something in 2004.
Re: note 2
Could this be an artifact of the mid-decade redistricting that Texas and I believe Georgia experienced that affected the 2004 election? So in a way, you may be seeing the result of partisan gerrymandering- the last time it spiked was when we clearly had it. Remember, even during as far back as 94 and 96, lines were still moving in Texas, that time because of D controlled redrawing (and the advent of the technology that allowed you to do it).
It’s interesting that the differences within the 2000s decade are comparable with the jump from 2010 to 2012, and dwarf the not-much-of-a-jump from 2000 to 2002. Apparently there’s something more at play there besides gerrymandering (and probably besides incumbency too).
One suggestion – maybe you can plot the difference in the percentage needed for a 218/218 house against the number of D/R seats in the House before the election. That could give a measure of exactly how much of the difference is explained by incumbency.
Another data point to look at, of course, is the number of state legislatures fully controlled by each party when redistricting is done.
In addition to the D% needed to control, it would be interesting to put a line on the same graph that shows the actual result from the election. That would tie in with the changes in color of the bottom bars, and would highlight instances (like 2012) where one party got a majority of the vote in House races without getting control in the House.
I was under the impression that redistricting is only supposed to take place every 10 years, in the wake of a census. It strikes me odd that they managed to pass a really right-skewed redistricting plan in tx in 2006 and another (on a federal and state holiday) in VA this year and get away with it.
Dave, there’s no law against it. Just a convention of normal governance that we don’t change districts more often. It’s bad news when such norm-based mechanisms break down.
It seems like the thing you really want to show is the derivative / Delta vote needed. There is clearly a structural advantage to being an incumbent. And that is built in to the percentage needed. That’s why things are relatively flat over the periods of one party rule.
Big changes in the percentage indicate something happened to shift the playing field. That might be policy (Republican revolution in 94) or other external factors like gerrymandering.
Theoretically you’d expect policy changes to affect the actual distribution of votes between the parties, but NOT the percentage of the votes needed to gain a majority. So, let’s say, a Dem wave year should produce a large popular vote victory for the Dems, but not shift the size of victory needed to gain control of the House into negative territory.
I’m not certain what the fact is that the graph is communicating (which is a question for the designer, but also about what you’re trying to say). Are you trying to demonstrate an overall shift in vote share that would have been necessary to swing the house (like your meta-margin), or the amount that would lead to 218 dems? Either way it might be interesting to see the actual vote share and the actual outcome along with this calculation to have some more context about what specific point you’re trying to make.
It appears that the structural advantage always lies in the direction of the winning party, which suggests that this may be measuring something other than a geographical factor independent of the election itself. Can you plot actual D% vs. D% needed for control? It seems likely to be very correlated, which would be odd if this were a measure of purely districting information.
The election-to-election fluctuations do not appear to identify redistricting years either, which also suggests this is not a measure of districting effects.
Or it could suggest demographic and migrational effects so large during the course of the 2000’s that they dwarf the impact of districting. I’d be pretty surprised by that though, since 2008 to 2010 would seem to refute that.
In the run-up to the election you had redistricting as a 2.5 advantage to Rs. I think for this to be truly useful you should localize it. Then Organizing for Action could use it for goal setting.
🙂
Like you localized fundage allocation!